On Thu, 16 Sep 2021 10:33:09 GMT, Daniel Fuchs <dfu...@openjdk.org> wrote:
>> I pushed a fix to run the test in othervm mode. Please test again.
>
> Meanwhile I did some experiment and was puzzled by the fact that the new test
> still failed intermittently - while the corresponding
> SctpMultiChannel/CloseDescriptors did not. By comparing the two I noticed
> that the SctpMultiChannel test is introducing some delays at key places. I
> did the same to the new test - and behold! It stopped failing.
>
> Here are my changes (ignore the /othervm which you already did):
>
>
> diff --git a/test/jdk/com/sun/nio/sctp/SctpChannel/CloseDescriptors.java
> b/test/jdk/com/sun/nio/sctp/SctpChannel/CloseDescriptors.java
> index 99f9563e016..11e2e52f577 100644
> --- a/test/jdk/com/sun/nio/sctp/SctpChannel/CloseDescriptors.java
> +++ b/test/jdk/com/sun/nio/sctp/SctpChannel/CloseDescriptors.java
> @@ -26,7 +26,7 @@
> * @bug 8238274
> * @summary Potential leak file descriptor for SCTP
> * @requires (os.family == "linux")
> - * @run main CloseDescriptors
> + * @run main/othervm CloseDescriptors
> */
>
> import java.io.BufferedReader;
> @@ -79,9 +79,12 @@ public class CloseDescriptors {
> selThread = new SelectorThread();
> selThread.start();
>
> + // give time for the server and selector to start
> + Thread.sleep(100);
> for (int i = 0 ; i < 100 ; ++i) {
> System.out.println(i);
> doIt(port);
> + Thread.sleep(100);
> }
> System.out.println("end");
> if (!check()) {
> @@ -110,6 +113,7 @@ public class CloseDescriptors {
> catch (Exception ex) {
> ex.printStackTrace();
> }
> + Thread.sleep(200);
> }
> }
>
> @@ -199,4 +203,3 @@ public class CloseDescriptors {
> }
> }
> }
Thank you for your help. I applied your suggested fix and it works fine. But it
takes about 4 minutes to be finished. So I added a timeout parameter to 250.
-------------
PR: https://git.openjdk.java.net/jdk/pull/5274
